metadata
dataset_info:
features:
- name: image
dtype: image
- name: messages
list:
- name: question
dtype: string
- name: answer
dtype: string
- name: thought
dtype: string
- name: bbox
dtype: string
- name: metadata
struct:
- name: website
dtype: string
- name: url
dtype: string
- name: image_w
dtype: int32
- name: image_h
dtype: int32
splits:
- name: train
num_bytes: 1495075856
num_examples: 5000
download_size: 1495371868
dataset_size: 1495075856
configs:
- config_name: default
data_files:
- split: train
path: data/train-*